In South Africa, there is an extreme paucity of information relating maize (Zea mays L.) yields to soil and plant K levels and most K fertilizer recommendations are highly subjective. Controversy also surrounds the practice of relating K fertilizer recommendations to estimates of yield potential. The work reported here was conducted to address these shortcomings. The yield and plant compositional response of maize to a range of soil K levels was studied for eight seasons on a Metz clay soil (clayey, kaolinitic, thermic Plinthustalf). Relative yields were related to exchangeable soil K and the leaf parameters: K content, (Ca + Mg)/K and K ORIS index. Although grain yield maxima varied from 5.85 to 10.80 Mg ha-1 , 91 % of the relative yield variability was explainable in terms of exchangeable soil K. Leaf-based indices were also closely related to relative yield, but were sensi!ive to variatio~s in top~oil moisture status prior .to sampling ~nd ~ere ~~t considered dia~nosti cally superior to the K sOil test. It IS concluded that the optimum K level In thiS SOil IS c. 0.32 cmol L-, that there is no justification for relating K requirement to estimates of yield potential and that leaf diagnostic criteria should be interpreted with circumspection.In Suid Afrika bestaan daar 'n groot gebrek aan inligting betreffende die verband tussen mielie (Zea mays L.) opbrengste en grond en plant K peile en die meeste K bemestings -aanbevelings is dan ook hoogs subjektief. Daar bestaan ook verwarring rondom die koppeling van K bemestings -aanbevelings aan skattings van opbrengs potensiaal. Hierdie werk is onderneem ten einde die tekortkomings aan te spreek. Die opbrengs en plant samestelling respons van mielies onderworpe aan 'n reeks van grond K peile is bestudeer oor 'n tydperk van agt seisoene op 'n Metz kleigrond (kleierige, kaolinitiese, termiese Plinthustalf). Relatiewe opbrengste is in verband gebring met uitruilbare grond K en die blaar parameters, K-inhoud, (Ca + Mg)/K en ORIS indeks. Alhoewel die maksimum graanopbrongs gevarieer het van 5.85 tot 10.80 Mg ha-1 , was 91 % van die relatiewe opbrengs variansie verklaarbaar in terme van uitruilbare grond K. Blaar-gebaseerde indekse was ook nou verwant aan relatiewe opbrengs, maar was sensitief tot veranderings in bogrond vog status voor monsterneeming en was nie as diagnosties beter as grond K toets beskou nie. Die gevolgtrekking van hierdie studie is dat die optimum grond K peil in hierdie grond ongeveer 0.32 cmol L-1 is en dat daar geen regverdiging bestaan om K benodighede te koppel aan skattings van opbrengs potensiaal nie. Blaar diagnostiese norme moet met omsigtigheid ge"interpreteerword.